Adaptive Step Frequency Detection and Stride Length Estimation for Pedestrian Dead Reckoning Based on Wearable Devices

Pedestrian Dead Reckoning (PDR) systems are widely used in indoor positioning due to their minimal environmental impact, low cost, and ease of deployment. Step frequency detection and stride length estimation are two key components of PDR systems, and detecting stride length under different walking modes presents a complex challenge. To address this issue, this paper proposes a method for recognizing six walking modes (Walking, Running, Reverse Walking, Marching in Place, Right Lateral Walking and Left Lateral Walking) using wearable devices for both step frequency and stride length . Firstly, a combination of peak detection and zero-crossing detection algorithms is employed to enhance the accuracy of step frequency detection. Secondly, features extracted from each step are used with a Convolutional Neural Network (CNN) algorithm to classify and correct the six walking modes. Lastly, an improved adaptive stride length estimation algorithm is proposed for different walking modes to achieve higher accuracy in stride length estimation. Experimental results show that the step frequency accuracy exceeds 98%, the gait mode recognition accuracy is 95.33%, and the relative error in estimated walking distance is approximately 1%, outperforming existing common methods for step frequency and stride length estimation.